Lethal Pancreatitis

Brian P. Buggy,* M.D. and Timothy T. Nostrant, M.D. Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan

Thirty-two patients died of pancreatitis and its com- of pancreatitis and whether certain prognostic factors plications over a 10-year period. Infection (bacteremia, were useful in predicting the outcome ofthe disease. fungemia, or pancreatic ahscess) was the major cause of death in 80%. In the remaining 20%, refractory METHODS AND PATIENTS hypotension or respiratory failure were the lethal mechanisms. In only 78% of patients was the correct diag- The hospital records of patients who died with acute nosis made hefore death. Ninety-four percent of those pancreatitis were reviewed. Patients were included for who died did so during their first clinical episode of evaluation when pancreatitis was the predominant pancreatitis. Prophylactic antibiotics did not prevent cause of death and when a diagnosis of either diffuse the development of pancreatic abscesses and organisms hemorrhagic or necrotizing pancreatitis was made at resistant to the antibiotics used often became the pri- either surgery or autopsy. Patients were excluded when mary pathogens. Certain prognostic factors reliably sep- pancreatitis was not a significant finding at surgery or arated those who died from those who lived. Peritoneal autopsy. From 1969-1979 there were 32 patients at the lavage and dialysis may be helpful in both the early University of Michigan Hospitals in whom pancreatitis diagnosis and therapy of severe acute pancreatitis. was the predominant cause of death. Over the same 10- year period, 410 patients had a primary or secondary INTRODUCTION discharge diagnosis of acute pancreatitis for a mortality rate of 7.8%. An age, sex, and hospital service (medical "Acute pancreatitis is the most terrible of all the or surgical)-matched set of 32 patients who had an calamities that occur in connection with the abdominal episode of acute pancreatitis and lived was also studied. viscera. The suddenness of its onset, the illimitable agony which accompanies it, and the mortality attend- RESULTS ant upon it all render it the most formidable of catas- trophes" (1). In the group of patients who lived, there were 18 men Acute pancreatitis describes a clinical spectrum rang- and 14 women with a mean age of 39 years (range 17- ing from mild, self-limited symptoms of epigastric pain, 76). The average length of hospitalization was 9 days. nausea, and vomiting to an overwhelming, rapidly fatal Of those who died, there were 18 men and 14 women illness. Biliary tract disease and alcoholism remain the with a mean age of 49 (range 18-89). The diagnosis of most commonly recognized etiologies (2-4). Biliary pancreatitis was confirmed at autopsy in 75% and at tract disease has a mortality of 10-15% while alcoholic celiotomy in 25%. Diagnosis was made before death in pancreatitis has a 1-5% mortality rate (2-6). Mortality 78% and at autopsy in the remaining 22%. The average is highest (20-40%) in cases occurring postoperatively length of hospitalization before death was 51 days. In and in those in which there is no apparent etiology. both groups approximately two-thirds of the patients Overall, a given episode of pancreatitis has an approx- were on a surgical service and one-third on a medical imate 5-15% mortality, a figure which has remained service. The etiological groups are shown in Table 1. constant in most reviews over the last 20 years (7-10). Serum amyiase was measured in 90% ofthe patients This study was undertaken to examine the clinical who died. In 20% of these it was normal when first course of patients dying with acute pancreatitis. Data drawn. In both groups of patients the height of the were analyzed to assess the modes of presentation, the initial amyiase was >1000 IU/1 in 16 (Table 1). Seven role of concurrent disease in affecting survival, the of these instances were in patients with biliary tract actual mechanisms leading to death, the use of anti- disease while only two were in alcoholics. Although the biotics and their relation to the infectious complications height of the initial amyiase may correlate with the presence of cholelithiasis (11), it has never been shown * Present address: Department of Medicine, Medical College of to correlate with prognosis (4, 12, 13). Wisconsin and St. Luke's Hospital, 2900 West Oklahoma Avenue None ofthe patients who lived but one-third of those Milwaukee, WI 53215. who died were comatose on admission. Seven of these 810 December 1983 LETHAL PANCREATITIS 811

TABLE 1 TABLE 2 Clinical Summary Isolates from Pancrealic Abscesses (10 Patients)*

Previous E. coli 4 No. of Pa- Amyiase Clinical Pan- Pseudomonas aeruginosa 3 tients >1000 Etiological Group creatitis Klebsiella penumoniae 3 Enterobacter spp. 2 Lived Died Lived Died Lived Died Acinetobacter oxidans 2 1. Biliary disease 10 8 4 0 5 2 Serralia marcescens 2 II. Alcoholism 10 6 7 0 2 0 Proteus spp. 1 III. Postoperative 1 6 0 0 1 2 Citrobacter spp. 1 IV. Idiopathic 9 4 3 0 3 0 Streptococcus fecatis 6 V. Miscellaneous* 2 8 0 2 0 1 Streptococcus viridans I Staphylococcus aureus 1 Total 32 32 14 2 II 5 "Anaerobes" I C. albicans 3 * Miscellaneous causes included severe burns, gunshot wounds, polyarteritis nodosa, systemic lupus erythematosus, hypertriglyceri- * The abscesses were polymicrobial in nine patients. demia, use of L-asparaginase, and cold water drowning. were given antibiotics. Virtually every antibiotic on the 10 patients were also severely hypotensive. The pres- market during the years of the study was used. The ence of either an elevated serum amyiase or peritoneal most commonly used antibiotics were ampicillin or signs (rigid abdomen or absent bowel sounds) was cephalothin, usually in combination with gentamicin. helpful in localizing pathology to the abdomen in nine. In the group who died receiving antibiotics empirically, In the entire series, of the seven patients who did not all but one still died of an infectious complication and have the diagnosis made before death, three were un- all of these were with organisms resistant to the original conscious on admission. In one of them an amyiase antibiotic choice. Superinfection with Candida albicans was not measured; in one it was normal; and in one it occurred commonly during antibiotic therapy, being was elevated and apparently ignored. seen as part ofthe abscess flora in three patients and in Death occurred in the settings of infection, respira- the blood of six. In the patients who lived, no fungal tory failure, or renal failure. Infection played a domi- superinfection occurred. nant role in the death of 26 (82%) ofthe 32 patients. An adult respiratory distress syndrome-like picture Fifteen of these had pancreatic abscesses (diagnosed an was seen in almost 20% ofthe total group. It occurred average of 18 days after presentation) and the remainder in patients among all subgroups. The mean time of developed either overwhelming bacteremia or funge- onset of this complication was 3 days after symptoms mia. Subphrenic abscesses were seen in five patients, began and no cases appeared after 7 days. Arterial three of whom also had pancreatic abscesses. Most hypoxemia (pO2 <60mm Hg) was present in 50% of all patients with pancreatic abscesses were also bacteremic patients who died within the 1st wk of their illness. The with an organism ultimately found in their abscess. six patients who died of adult respiratory distress syn- Culture results were available for 10 of these patients drome all had either acute renal failure or a pancreatic (Table 2). All but one of these abscesses were polymi- abscess; in fact, half of these six had all three conditions. crobial. Anaerobes were isolated in only one patient Thus the lethal respiratory failure seen in this series but few anaerobic cultures were obtained. In patients occurred only with another major complication of pan- with cholelithiasis and postoperative pancreatitis who creatitis. died, infection was universally present. Twelve of these Renal failure was coexistent with the onset of pan- 14 patients had pancreatic abscesses and the remaining creatitis in four patients and developed in nine of the two had bacteremia. Infection was the cause of death patients who died. In the group with pre-existing renal in half of the alcoholic patients and in 75% of patients failure, one each also had accompanying biliary tract with miscellaneous or idiopathic causes for their pan- disease, alcoholism, or SLE. The remaining patient had creatitis. Only one infectious complication was seen in chronic pyelonephritis. Hypotension as an etiology for the patients who lived (a pancreatic abscess developing the renal failure could be clearly demonstrated in six of after cholecystectomy which was drained without com- these nine patients. plication). Several sets of criteria exist which attempt to predict Antibiotics were used in 30/32 (94%) ofthe patients the clinical outcome (survival or death) of an attack of who died and in 6/21 (29%) of those who lived. They acute pancreatitis based on an assessment of certain were used in a "prophylactic" sense (empiric therapy factors early in the course of the disease. Perhaps the on admission without documentation of infection) in best studied are those of Ranson and colleagues (4, 12- more than one-half of the patients in both groups who 14) who evaluated 43 objective findings during the first 812 BUGGY AND NOSTRANT Vol. 78, No. 12, 1983 48 h of pancreatitis and found 11 predictive of "serious" half of the deaths occur in the first few days, usually of pancreatitis; i.e., leading to death or requiring >7 days hypovolemia or respiratory complications. Those who in an intensive care unit (Table 3). In 350 patients with survive for 1 wk often succumb later to infectious <3 prognostic signs, the mortality was 0.9%. In 100 sequelae. From 10-50% of cases are diagnosed on the patients with >3 prognostic signs, mortality increased autopsy table and almost all of these are patients who from 16% in those with three or four signs to 100% in die within 24-48 h after admission (15-18). These are those with seven or eight signs (14). Factors shown to also the patients most likely to be comatose and have be not predictive of serious illness or mortality were the diagnosis missed (19). any findings on physical exam or the presence and Infection is the overwhelming cause of death in pa- height of the serum amyiase. In this series the mean tients surviving beyond 1 wk. A pancreatic abscess is number of prognostic factors present in those who died seen in 4-9% of patients with acute pancreatitis with a was 6.1 (range of 3-9) and in the group who lived a mortality rate of 25-57%, despite advances in diagnosis, mean of 0.3 (range 0-2; p < 0.001 by x^). Twenty of aggressive surgical management, and antibiotic therapy the patients who lived had none of these signs present. (20-23). Possible routes of infection include invasion In the group who died the most common signs were from infected bile, direct penetration through the wall leukocytosis, hypocalcemia, hypoxemia, and decreased ofthe transverse colon, and hematogenous or lymphatic renal function. spread (2). In approximately half, the abscesses are Therapeutic peritoneal dialysis was attempted in five polymicrobial although in this study 10/11 (91%) had of the patients who died. In three patients dialysis was multiple organisms. Most series were compiled before administered for renal failure late in their clinical the advent of current anaerobic technology but one course. Two patients received dialysis as therapy for report found Bacteroides spp. in 15% of cases (22). pancreatitis: one received it 7 days after onset and died Enterobacteriaceae predominate with Eseherichia coli 24 h later, and another died shortly after one cycle of and Klebsiella seen most frequently. C. albicans is seen dialysis had been given. only in well-established abscesses and probably repre- sents superinfection due to prior antibiotic therapy. DISCUSSION The controversy over the use of prophylactic anti- Previous studies of lethal pancreatitis, comprising biotics during acute severe pancreatitis has not been 435 cases studied with autopsy since the detailed path- settled. One retrospective (24) and three prospective ologic study of Fitz in 1889 (15), have underscored studies comprising over 700 patients have compared several facts observed in this study. The vast majority outcome in patients given prophylactic ampicillin (25, of patients who die of acute pancreatitis do so during 26) or cephalothin (24, 27) and found no difference in their first clinical episode ofthe disease. Of 315 patients morbidity or mortality. However, virtually all of the in the literature for which these data are obtainable patients in these studies were alcoholics with mild pan- (including the current series of 32), 17 (6%) had had a creatitis, a group expected to have the lowest mortality. prior clinical attack of pancreatitis. Approximately one- Should a pancreatic abscess develop, mortality is highest (80-85%) in those patients with biliary tract disease or postoperative etiologies (6). Most patients in TABLE 3 Ranson Criteria these groups, however, are already receiving antibiotics. Indeed, in the patients who died, half were already Died receiving antibiotics at the time that an ultimately fatal Lived (30 Evaluable) (12 infectious complication developed. Organisms (bacte- Evaluable) rial or fungal) that were subsequently recovered from On admission the blood or abscesses in these patients were uniformly Age >55 yr 5 9 resistant to the antibiotics initially used. The subgroup Glucose >200 mg/100 ml 0 12 with biliary tract disease may be one in which prophy- White blood cells >16,000/mm' 3 16 Lactose dehydrogenase >35O IU/1 0 10 lactic antibiotics would deserve a trial, since their mor- SGOT >250 Sigma-Frankel U/100 ml 1 4 tality is so high if infection does occur. At present, however, there is no good data to support the use of Within 48 h prophylactic antibiotics in acute pancreatitis, especially Decreased hematocrit by >10% 1 13 in patients who are not critically ill on presentation. Calcium <8 mg/100 ml 4 14 Base deficit >4 mEg/1 0 12 The adult respiratory distress syndrome occurred in Blood urea nitrogen increase by >5 mg/ 0 17 20% of those dying. All six of these patients died, which 100 ml is a higher mortality from this process than that previ- Fluid sequestration >6 1 0 11 ously reported (28). However, a similarly high mortality PO2 <60 mm Hg 0 16 from respiratory failure in alcoholics with acute pan- December 1983 LETHAL PANCREATITIS 813 creatitis has been observed (16, 27). These six patients in unconscious patients, as their pancreatitis is often all had at least one other major metabolic process, such diagnosed postmortem (19). The role of prophylactic as renal failure or infection, which may have directly antibiotics has not been clarified in severely ill patients, contributed to this higher mortality. but if proven beneficial, may offer a means of dimin- Pancreatitis has been reported to occur at a higher ishing the late mortality from infection (especially in rate in patients with pre-existing renal failure (3). In patients with biliary tract pathology). The role of peri- this series, however, three of four such patients had toneal lavage and dialysis in the early diagnosis and another associated condition such as alcoholism. Oli- therapy of severe pancreatitis deserves further study. guric renal failure is not uncommon in acute pancreatitis, especially in the first 24-48 h after presentation. Reprint requests: Dr. Timothy T. Nostrant., Department of Inter- nal Medicine, D-2105 South Ambulatory Care Building, University Hypovolemia as a cause of renal failure could account ofMichigan, Ann Arbor, MI 48109. for six of nine such instances in this study. However, in three of these nine and in at least 18 other patients described in the literature (29, 30), hypotension could REFERENCES not be documented (although perhaps not detected) 1. Moynihan B. Acute pancreatitis. Ann Surg 1925:81:132-42. and some other factor, such as a circulating vasopressor 2. Banks PA. Pancreatitis. New York: Plenum Medical Book Com- pany, 1979. having selective renal activity, has been postulated to 3. Toskes PP, Greenberger NJ. Acute and chronic pancreatitis. In: cause acute renal failure (2, 31). Cotsonas NJ, ed. Disease-a-month. Vol 29. No 6. Chicago: Year Several attempts have been made to predict the clin- Book Medical Publishers, 1983:1-81. 4. Ranson JHC. Acute pancreatitis. In: Ravitch MM, ed. Current ical severity of a given episode of acute pancreatitis. problems in surgery. Vol 16. No 11. Chicago: Year Book Medical The studies of Ranson and colleagues (4, 12-14) did Publishers, 1979:1-83. not fmd predictive utility from any parameters on 5. Frey CF. The operative treatment of pancreatitis. Arch Surg 1969,98:406-17. physical exam, while others have (32). This study also 6. Gliedman ML, Bolooki H, Rosen RG. Acute pancreatitis. In: illustrated the use of Ranson's criteria in predicting a Ravitch MM, ed. Current problems in surgery, Vol 7. No 8. high risk group of patients with those living having a Chicago: Year Book Medical Publishers, 1970:1-52. 7. Welbourn RB, Armitage P, Gilmore OJA, et al. Death from mean of 0.5 prognostic factors while those dying had acute pancreatitis. MRC multicentre trial of glucagon and apro- 6.1. tonin. Lancet 1977;2:632-5. 8. Olsen H. Pancreatitis. A prospective clinical evaluation of 100 There has been recent interest in the role of peritoneal cases and review ofthe literature. Am J Digest Dis 1974; 19:1077- lavage and dialysis in both the diagnosis and therapy of 90. acute pancreatitis early in its course. Diagnostic peri- 9. Lukash WM. Complications of acute pancreatitis. Arch Surg 1967;94:848-52. toneal lavage (considered positive if more than 10 ml 10. Albo R, Silen W, Goldman L. A critical clinical analysis of acute of free peritoneal fluid was obtained or any lavage pancreatitis. Arch Surg 1963,86:174-80. return with a dark color) was found as predictive ofthe 11. Scholhamer CF, Spiro H M. The first attack of acute pancreatitis: a clinical study. J Clin Gastroenteroi 1979; 1:325-9. subsequent clinical course as a clinical grading scheme 12. Ranson JHC, Rifkind KM, Roses DF, et al. Objective early or the criteria of McMahon et al. (33). A randomized identification of severe acute pancreatitis. Am J Gastroenteroi trial of therapeutic peritoneal lavage in 70 patients 1974;61:443-51. 13. Ranson JHC, Pasternack BS. Statistical methods for quantifying thought to have severe alcoholic pancreatitis (hypocal- the severity of clinical acute pancreatitis. J Surg Res I977;22:79- cemia, acidosis, hypotension, hypoxemia) found "im- 91. provement" at 24 h and less mortality in the group 14. Ranson JHC. Acute pancreatitis. Surg Clin North Am dialyzed (27). An earlier report of Ranson and Spencer 1981;61:55-70. 15. Fitz RH. Acute pancreatitis. Boston Med Surg J 1889;120:181- (34) showed a dramatic reduction in the early mortality 207, 229-235. of pancreatitis in patients given 48 h of peritoneal 16. Peterson LM, Brooks JR. Lethal pancreatitis: a diagnostic di- dialysis. However, late mortality (almost all due to lemna. AmJ Surg 1979; 137:491-6. 17. Shader AE, Paxton JR. Fatal pancreatitis. Am J Surg pancreatic abscesses) was unaffected. These studies 1966;111:369-73. have thus suggested at least improved early survival in 18. Storck G, Pettersson G, Edlund Y. A study of autopsies upon those patients supported with peritoneal dialysis and it 116 patients with acute pancreatitis. Surg Gynecol Obstet 1976;143:241-5. may deserve a wider role in diagnosis and therapy than 19. Tofner AH, Spiro HM. Shock or coma as the predominant it currently has. manifestation of painless acute pancreatitis. Ann Intern Med Since virtually all patients who die of acute pancrea- 1962;57:655-9. 20. Frey CF, Lindenauer SM. Pancreatic abscess. In: Dent TL, titis do so during their first clinical episode ofthe disease Eckhauser FE, Vinik AI, et ah, eds. Pancreatic disease—diagnosis (regardless of etiological subgroup), particular attention and therapy. New York: Grune & Stratton, 1981:257-84. to their changing physiology and the appearance of 21. Evans FC. Pancreatic abscess. Am J Surg 1969;! 17:537-40. 22. Holden JL, Berne TV, Rosoff L. Pancreatic abscess following prognostic factors should be assessed early in the clinical acute pancreatitis. Arch Surg 1976;111:858-61. course. Serum amyiase, although not predictive, is a 23. Ranson JHC, Spencer FC. Prevention, diagnosis and treatment useful screening device and may have particular utility of pancreatic abscess. Surgery 1977;82:227-34. 814 BUGGY AND NOSTRANT Vol. 78, No. 12, 1983 24. Kodesch R, DuPont HL. Infectious complications of acute pan- J 1972;2:801-2. creatitis. Surg. Gynecol Obstet 1973; 136:763-8. 30. Goldstein DA, Llach F, Massry SG. Acute renal failure in patients 25. Howes R, Zuidema GD, Cameron JL. Evaluation of prophylactic with acute pancreatitis. Arch Intern Med 1976; 136:1363-5. antibiotics in acute pancreatitis. J Surg Res 1975; 18:197-200. 31. Werner MH, Hayes DF, Lucas CE, et al. Renal vasoconstriction 26. Finch WT, Sawyers JL, Schenker S. A prospective study to in association with acute pancreatitis. Am J Sure 1974-127-185- determine the efficacy of antibiotics in acute pancreatitis. Ann 90. Surg. 1976; 183:667-71. 32. Jacobs ML, Daggett WM, Civetta JM, et al. Acute pancreatitis: 27. Stone HH, Fabian TC. Peritoneal dialysis in the treatment of analysis of factors influencing survival. Ann Surg 1977;185:43- acute alcoholic pancreatitis. Surg Gynecol Obstet 1980; 150-878- 82. 33. McMahon MJ, Playforth MJ, Pickford IR. A comparative study 28. Warshaw AL, Lesser PB, Rie M, et al. The pathogenesis of of methods for the prediction of severity of attacks of acute pulmonary edema in acute pancreatitis. Ann Surg 1975; 182:505- pancreatitis. Br J Surg 1980;67:22-5. 10. 34. Ranson JHC, Spencer FC. The role of peritoneal lavage in severe 29. Gordon D, Calne RY. Renal failure in acute pancreatitis. Br Med acute pancreatitis. Ann Surg 1978; 187:565-75.